Create app.py

app.py

@@ -0,0 +1,78 @@
+import chromadb
+import gradio as gr
+from sentence_transformers import SentenceTransformer
+from llama_cpp import Llama
+
+# ✅ Initialize ChromaDB
+chroma_client = chromadb.PersistentClient(path="./chromadb_store")
+collection = chroma_client.get_or_create_collection(name="curly_strings_knowledge")
+
+# ✅ Load Local Embedding Model
+embedder = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
+
+# ✅ Load Fine-Tuned LLaMA Model
+llm = Llama.from_pretrained(
+    repo_id="krishna195/second_guff",
+    filename="unsloth.Q4_K_M.gguf",
+)
+
+# ✅ File-Based Search Function
+def search_in_file(query, file_path="merged_output.txt"):
+    try:
+        with open(file_path, "r", encoding="utf-8") as file:
+            lines = file.readlines()
+        
+        # Search for the query in file content
+        matched_lines = [line.strip() for line in lines if query.lower() in line.lower()]
+        
+        return "\n".join(matched_lines) if matched_lines else "No relevant data found in file."
+    
+    except FileNotFoundError:
+        return "File not found. Please check the file path."
+
+# ✅ Retrieve Context from ChromaDB & File
+def retrieve_context(query):
+    query_embedding = embedder.encode(query).tolist()
+    results = collection.query(query_embeddings=[query_embedding], n_results=2)
+
+    retrieved_texts = [doc for sublist in results.get("documents", []) for doc in sublist if isinstance(doc, str)]
+    
+    # If no result from ChromaDB, try searching in the file
+    if not retrieved_texts:
+        return search_in_file(query)
+    
+    return "\n".join(retrieved_texts)
+
+# ✅ Chatbot Function with Optimized Retrieval
+def chatbot_response(user_input):
+    context = retrieve_context(user_input)
+
+    messages = [
+        {"role": "system", "content": """You are an expert on the Estonian folk band Curly Strings.  
+        - Use the **retrieved knowledge** from ChromaDB or the file to answer.  
+        - If a **song** is mentioned, provide its name and **suggest similar tracks**.  
+        - If no match is found, say "I couldn’t find details, but here’s what I know."."""},
+        {"role": "user", "content": user_input},
+        {"role": "assistant", "content": f"Retrieved Context:\n{context}"},
+    ]
+
+    response = llm.create_chat_completion(
+        messages=messages,
+        temperature=0.4,  
+        max_tokens=300,
+        top_p=0.9,
+        frequency_penalty=0.7,
+    )
+
+    return response["choices"][0]["message"]["content"].strip()
+
+# ✅ Gradio Chatbot Interface
+iface = gr.Interface(
+    fn=chatbot_response,
+    inputs=gr.Textbox(label="Ask me about Curly Strings 🎻"),
+    outputs=gr.Textbox(label="Bot Response 🎶"),
+    title="Curly Strings Chatbot",
+    description="Ask about the Estonian folk band Curly Strings! Now also searches in 'merged_output.txt'.",
+)
+
+iface.launch()